Inference of Pavement Properties with Roadside Accelerometers

An array of four synchronized single-axis accelerometers was fixed to the surface of an asphalt pavement. Vertical acceleration traces triggered by several nearby passes of a truck with known characteristics were recorded. The work focused on presenting and demonstrating an interpretation method for inferring the mechanical properties of the pavement system based on the recorded accelerations. In general terms, the method was based on careful low-pass filtering the field-measured acceleration traces, and then best-matching them with a corresponding set of calculated acceleration traces. For this purpose, the pavement system was modeled as a two-layered linear elastic half-space, and a model-guided signal filtering approach was devised to ensure that irrelevant signal content is removed prior to the matching. Based on the analysis of six separate truck passes it was noticed that the inferred upper layer modulus exhibited medium variability (coefficient of variation of 45%) while the lower (subgrade) modulus showed little variability (coefficient of variation of 8%). The moduli values displayed fair agreement with those independently estimated from non-destructive and semi-destructive tests. By analyzing many more passes inferred moduli are expected to become more representative. Overall, the method seems workable and scalable, with capacity to handle any number of acceleration sensors as well as other sensor types.